Skip to content

Instantly share code, notes, and snippets.

@wboykinm
Forked from kerryrodden/.block
Last active August 29, 2015 13:56
Show Gist options
  • Star 1 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save wboykinm/9078694 to your computer and use it in GitHub Desktop.
Save wboykinm/9078694 to your computer and use it in GitHub Desktop.

This example shows how it is possible to use a D3 sunburst visualization (partition layout) with data that describes sequences of events.

A good use case is to summarize navigation paths through a web site, as in the sample synthetic data file (visit_sequences.csv). The visualization makes it easy to understand visits that start directly on a product page (e.g. after landing there from a search engine), compared to visits where users arrive on the site's home page and navigate from there. Where a funnel lets you understand a single pre-selected path, this allows you to see all possible paths.

Features:

  • works with data that is in a CSV format (you don't need to pre-generate a hierarchical JSON file, unless your data file is very large)
  • interactive breadcrumb trail helps to emphasize the sequence, so that it is easy for a first-time user to understand what they are seeing
  • percentages are shown explicitly, to help overcome the distortion of the data that occurs when using a radial presentation

If you want to simply reuse this with your own data, here are some tips for generating the CSV file:

  • no header is required (but it's OK if one is present)
  • use a hyphen to separate the steps in the sequence
  • the step names should be one word only, and ideally should be kept short. Non-alphanumeric characters will probably cause problems (I haven't tested this).
  • every sequence should have an "end" marker as the last element, unless it has been truncated because it is longer than the maximum sequence length (6, in the example). The purpose of the "end" marker is to distinguish a true end point (e.g. the user left the site) from an end point that has been forced by truncation.
  • each line should be a complete path from root to leaf - don't include counts for intermediate steps. For example, include "home-search-end" and "home-search-product-end" but not "home-search" - the latter is computed by the partition layout, by adding up the counts of all the sequences with that prefix.
  • to keep the number of permutations low, use a small number of unique step names, and a small maximum sequence length. Larger numbers of either of these will lead to a very large CSV that will be slow to process (and therefore require pre-processing into hierarchical JSON).

I created this example in my work at Google, but it is not part of any Google product. It is covered by the Apache license:

Copyright 2013 Google Inc. All Rights Reserved.

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<title>Sequences sunburst</title>
<script src="http://d3js.org/d3.v3.min.js"></script>
<link rel="stylesheet" type="text/css"
href="https://fonts.googleapis.com/css?family=Open+Sans:400,600">
<link rel="stylesheet" type="text/css" href="sequences.css"/>
</head>
<body>
<div id="main">
<div id="sequence"></div>
<div id="chart">
<div id="explanation" style="visibility: hidden;">
<span id="percentage"></span><br/>
of visits begin with this sequence of pages
</div>
</div>
</div>
<div id="sidebar">
<input type="checkbox" id="togglelegend"> Legend<br/>
<div id="legend" style="visibility: hidden;"></div>
</div>
<script type="text/javascript" src="sequences.js"></script>
<script type="text/javascript">
// Hack to make this example display correctly in an iframe on bl.ocks.org
d3.select(self.frameElement).style("height", "700px");
</script>
</body>
</html>
pool-dropout-end 187
pool-audit-dropout-end 2218
pool-audit-success-end 1874
pool-simultaneous-end 13
pool-success-end 269
pool-success-audit-end 11
body {
font-family: 'Open Sans', sans-serif;
font-size: 12px;
font-weight: 400;
background-color: #fff;
width: 960px;
height: 700px;
margin-top: 10px;
}
#main {
float: left;
width: 750px;
}
#sidebar {
float: right;
width: 100px;
}
#sequence {
width: 600px;
height: 70px;
}
#legend {
padding: 10px 0 0 3px;
}
#sequence text, #legend text {
font-weight: 600;
fill: #fff;
}
#chart {
position: relative;
}
#chart path {
stroke: #fff;
}
#explanation {
position: absolute;
top: 260px;
left: 305px;
width: 140px;
text-align: center;
color: #666;
z-index: -1;
}
#percentage {
font-size: 2.5em;
}
// Dimensions of sunburst.
var width = 750;
var height = 600;
var radius = Math.min(width, height) / 2;
// Breadcrumb dimensions: width, height, spacing, width of tip/tail.
var b = {
w: 75, h: 30, s: 3, t: 10
};
// Mapping of step names to colors.
var colors = {
"home": "#5687d1",
"product": "#7b615c",
"search": "#de783b",
"account": "#6ab975",
"other": "#a173d1",
"end": "#bbbbbb"
};
// Total size of all segments; we set this later, after loading the data.
var totalSize = 0;
var vis = d3.select("#chart").append("svg:svg")
.attr("width", width)
.attr("height", height)
.append("svg:g")
.attr("id", "container")
.attr("transform", "translate(" + width / 2 + "," + height / 2 + ")");
var partition = d3.layout.partition()
.size([2 * Math.PI, radius * radius])
.value(function(d) { return d.size; });
var arc = d3.svg.arc()
.startAngle(function(d) { return d.x; })
.endAngle(function(d) { return d.x + d.dx; })
.innerRadius(function(d) { return Math.sqrt(d.y); })
.outerRadius(function(d) { return Math.sqrt(d.y + d.dy); });
// Use d3.text and d3.csv.parseRows so that we do not need to have a header
// row, and can receive the csv as an array of arrays.
d3.text("lifecycle.csv", function(text) {
var csv = d3.csv.parseRows(text);
var json = buildHierarchy(csv);
createVisualization(json);
});
// Main function to draw and set up the visualization, once we have the data.
function createVisualization(json) {
// Basic setup of page elements.
initializeBreadcrumbTrail();
drawLegend();
d3.select("#togglelegend").on("click", toggleLegend);
// Bounding circle underneath the sunburst, to make it easier to detect
// when the mouse leaves the parent g.
vis.append("svg:circle")
.attr("r", radius)
.style("opacity", 0);
// For efficiency, filter nodes to keep only those large enough to see.
var nodes = partition.nodes(json)
.filter(function(d) {
return (d.dx > 0.005); // 0.005 radians = 0.29 degrees
});
var path = vis.data([json]).selectAll("path")
.data(nodes)
.enter().append("svg:path")
.attr("display", function(d) { return d.depth ? null : "none"; })
.attr("d", arc)
.attr("fill-rule", "evenodd")
.style("fill", function(d) { return colors[d.name]; })
.style("opacity", 1)
.on("mouseover", mouseover);
// Add the mouseleave handler to the bounding circle.
d3.select("#container").on("mouseleave", mouseleave);
// Get total size of the tree = value of root node from partition.
totalSize = path.node().__data__.value;
};
// Fade all but the current sequence, and show it in the breadcrumb trail.
function mouseover(d) {
var percentage = (100 * d.value / totalSize).toPrecision(3);
var percentageString = percentage + "%";
if (percentage < 0.1) {
percentageString = "< 0.1%";
}
d3.select("#percentage")
.text(percentageString);
d3.select("#explanation")
.style("visibility", "");
var sequenceArray = getAncestors(d);
updateBreadcrumbs(sequenceArray, percentageString);
// Fade all the segments.
d3.selectAll("path")
.style("opacity", 0.3);
// Then highlight only those that are an ancestor of the current segment.
vis.selectAll("path")
.filter(function(node) {
return (sequenceArray.indexOf(node) >= 0);
})
.style("opacity", 1);
}
// Restore everything to full opacity when moving off the visualization.
function mouseleave(d) {
// Hide the breadcrumb trail
d3.select("#trail")
.style("visibility", "hidden");
// Deactivate all segments during transition.
d3.selectAll("path").on("mouseover", null);
// Transition each segment to full opacity and then reactivate it.
d3.selectAll("path")
.transition()
.duration(1000)
.style("opacity", 1)
.each("end", function() {
d3.select(this).on("mouseover", mouseover);
});
d3.select("#explanation")
.transition()
.duration(1000)
.style("visibility", "hidden");
}
// Given a node in a partition layout, return an array of all of its ancestor
// nodes, highest first, but excluding the root.
function getAncestors(node) {
var path = [];
var current = node;
while (current.parent) {
path.unshift(current);
current = current.parent;
}
return path;
}
function initializeBreadcrumbTrail() {
// Add the svg area.
var trail = d3.select("#sequence").append("svg:svg")
.attr("width", width)
.attr("height", 50)
.attr("id", "trail");
// Add the label at the end, for the percentage.
trail.append("svg:text")
.attr("id", "endlabel")
.style("fill", "#000");
}
// Generate a string that describes the points of a breadcrumb polygon.
function breadcrumbPoints(d, i) {
var points = [];
points.push("0,0");
points.push(b.w + ",0");
points.push(b.w + b.t + "," + (b.h / 2));
points.push(b.w + "," + b.h);
points.push("0," + b.h);
if (i > 0) { // Leftmost breadcrumb; don't include 6th vertex.
points.push(b.t + "," + (b.h / 2));
}
return points.join(" ");
}
// Update the breadcrumb trail to show the current sequence and percentage.
function updateBreadcrumbs(nodeArray, percentageString) {
// Data join; key function combines name and depth (= position in sequence).
var g = d3.select("#trail")
.selectAll("g")
.data(nodeArray, function(d) { return d.name + d.depth; });
// Add breadcrumb and label for entering nodes.
var entering = g.enter().append("svg:g");
entering.append("svg:polygon")
.attr("points", breadcrumbPoints)
.style("fill", function(d) { return colors[d.name]; });
entering.append("svg:text")
.attr("x", (b.w + b.t) / 2)
.attr("y", b.h / 2)
.attr("dy", "0.35em")
.attr("text-anchor", "middle")
.text(function(d) { return d.name; });
// Set position for entering and updating nodes.
g.attr("transform", function(d, i) {
return "translate(" + i * (b.w + b.s) + ", 0)";
});
// Remove exiting nodes.
g.exit().remove();
// Now move and update the percentage at the end.
d3.select("#trail").select("#endlabel")
.attr("x", (nodeArray.length + 0.5) * (b.w + b.s))
.attr("y", b.h / 2)
.attr("dy", "0.35em")
.attr("text-anchor", "middle")
.text(percentageString);
// Make the breadcrumb trail visible, if it's hidden.
d3.select("#trail")
.style("visibility", "");
}
function drawLegend() {
// Dimensions of legend item: width, height, spacing, radius of rounded rect.
var li = {
w: 75, h: 30, s: 3, r: 3
};
var legend = d3.select("#legend").append("svg:svg")
.attr("width", li.w)
.attr("height", d3.keys(colors).length * (li.h + li.s));
var g = legend.selectAll("g")
.data(d3.entries(colors))
.enter().append("svg:g")
.attr("transform", function(d, i) {
return "translate(0," + i * (li.h + li.s) + ")";
});
g.append("svg:rect")
.attr("rx", li.r)
.attr("ry", li.r)
.attr("width", li.w)
.attr("height", li.h)
.style("fill", function(d) { return d.value; });
g.append("svg:text")
.attr("x", li.w / 2)
.attr("y", li.h / 2)
.attr("dy", "0.35em")
.attr("text-anchor", "middle")
.text(function(d) { return d.key; });
}
function toggleLegend() {
var legend = d3.select("#legend");
if (legend.style("visibility") == "hidden") {
legend.style("visibility", "");
} else {
legend.style("visibility", "hidden");
}
}
// Take a 2-column CSV and transform it into a hierarchical structure suitable
// for a partition layout. The first column is a sequence of step names, from
// root to leaf, separated by hyphens. The second column is a count of how
// often that sequence occurred.
function buildHierarchy(csv) {
var root = {"name": "root", "children": []};
for (var i = 0; i < csv.length; i++) {
var sequence = csv[i][0];
var size = +csv[i][1];
if (isNaN(size)) { // e.g. if this is a header row
continue;
}
var parts = sequence.split("-");
var currentNode = root;
for (var j = 0; j < parts.length; j++) {
var children = currentNode["children"];
var nodeName = parts[j];
var childNode;
if (j + 1 < parts.length) {
// Not yet at the end of the sequence; move down the tree.
var foundChild = false;
for (var k = 0; k < children.length; k++) {
if (children[k]["name"] == nodeName) {
childNode = children[k];
foundChild = true;
break;
}
}
// If we don't already have a child node for this branch, create it.
if (!foundChild) {
childNode = {"name": nodeName, "children": []};
children.push(childNode);
}
currentNode = childNode;
} else {
// Reached the end of the sequence; create a leaf node.
childNode = {"name": nodeName, "size": size};
children.push(childNode);
}
}
}
return root;
};
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment